IT company

Building Blocks of AI : Machine Learning – Blog Profile

By Prosyscom
In March 12, 2018
112 Views


Building Blocks of AI : Machine Learning

If you have been following the tech lately, it’s certain that you must have come across the term ‘Machine Learning’ and ‘Deep Learning’. These terms are often misunderstood by many newbies in this field. Hence in this article I will try to explain in simple language the difference between them and how Machine Learning fits in AI. Machine learning is a lot like teenage ***. Everybody talks about it. Only some really know how to do it. Everyone thinks everyone else is doing it. So, everyone claims they’re doing it.  

UNDERSTANDING ‘MACHINE LEARNING’

Consider you are trying to toss a paper to a dustbin. After first attempt you realise that you have put too much force in it. After second attempt you realise you are closer to target but you need to increase your throw angle. What is happening here is basically after every throw we are learning something and improving the end result. We are programmed to learn from our experience.
We can do something similar with machines too. We can program a machine to learn from every attempts/experiences/data-points and then improve the outcome. Let’s see paper toss example in Machine and Non-Machine approach.
In our above example, a generic program would tell computer to measure the distance and angle and apply some pre-defined formula to calculate the force required. Now if you add a fan (wind force) to your setup, this program will continuously miss target and won’t learn anything from it’s failed attempt. To get the outcome right, you need to reprogram taking wind factor into your formula.
Now, for the same example a Machine Learning program would begin with a generic formula but after every attempt/experience refactor it’s formula. As the formula is continuously improved using more experiences (data points) the outcome too improved. You see these things into action around you in YouTube’s Video Recommendations and Facebook’s News Feed Content etc.

 A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.

WHY DOES ML MATTER?

For businesses, ML can can facilitate better decision-making in real-time, without the need for human intervention. Essentially, the technology is intuitive enough to “learn” from past events and adapt for better performance in the future based on that information. By leveraging these precise algorithms, organizations can better identify profitable opportunities and more effectively avoid unknown risks.

  • Banks are using to predict potential Loan Defaulters and Fraud Prevention (which is a hot topic in India currently), thus taking preventive measures.
  • Healthcare Industry is using it in numerous ways, the one which I found most useful was Identifying if a person has a disease just by uploading a picture of the related body part to a ML System which being highly trained over millions of images, predicts the possibility of the person having that disease. Eg., Breast Cancer.
  • Governments are using ML for Counterterrorism by identifying trends in social media can provide indications and warning for potential terrorist threats. It is also been used for Criminal Justice by uncovering trends that ultimately reduce the potential of bias in crime reporting, policing, bail, sentencing, and parole decisions.

TYPES OF MACHINE LEARNING ALGORITHMS

ML is broadly classified into 3 types : Supervised Learning, Unsupervised Learning and Reinforcement Learning.

Supervised Learning : Where a program is “trained” on a pre-defined dataset. Based off its training data the program can make accurate decisions when given new data. Example: Using a training set of human tagged positive, negative and neutral tweets to train a sentiment analysis classifier.
Unsupervised Learning : Where a program, given a dataset, can automatically find patterns and relationships in that dataset. Example: Analyzing a dataset of emails and automatically grouping related emails by topic with no prior knowledge or training which is also known as the practice of clustering.
Reinforcement Learning : This method aims at using observations gathered from the interaction with the environment to take actions that would maximize the reward or minimize the risk. These algorithms continuously learns from the environment in an iterative fashion. In the process, the itt learns from its experiences of the environment until it explores the full range of possible states.

Next article will be on PREDICTIVE ANALYTICS, stay tuned! 🙂 

قالب وردپرس